AMP Advances in Mathematical Physics 1687-9139 1687-9120 Hindawi Publishing Corporation 10.1155/2015/519374 519374 Research Article Infinitely Many Standing Waves for the Nonlinear Chern-Simons-Schrödinger Equations Seok Jinmyoung Mironov Andrei D. Department of Mathematics Kyonggi University 154-42 Gwanggyosan-ro Yeongtong-gu Suwon 443-760 Republic of Korea kyonggi.ac.kr 2015 1982015 2015 02 07 2015 05 08 2015 1982015 2015 Copyright © 2015 Jinmyoung Seok. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

We prove the existence of infinitely many solutions of the nonlinear Chern-Simons-Schrödinger equations under a wide class of nonlinearities. This class includes the standard power-type nonlinearity with exponent p>4. This extends the previous result which covers the exponent p>6.

1. Introduction

In [1, 2], Jackiw and Pi introduce a nonrelativistic model that the nonlinear Schrödinger dynamics is coupled with the Chern-Simons gauge terms as follows:(1)iD0ϕ+D1D1+D2D2ϕ=-λϕp-2ϕ,0A1-1A0=-Imϕ¯D2ϕ,0A2-2A0=Imϕ¯D1ϕ,1A2-2A1=-12ϕ2.Here, i denotes the imaginary unit, 0=/t, 1=/x1, 2=/x2 for (t,x1,x2)R1+2,   ϕ:R1+2C is a complex scalar field, Aμ:R1+2R is a component of gauge potential and Dμ=μ+iAμ is a covariant derivative for μ running over 0,1,2, and λ>0 is a parameter. The Chern-Simons gauge theory appears in the 1980s to explain electromagnetic phenomena of anyon physics such as the high temperature superconductivity or the fractional quantum Hall effect. In this paper, we are interested in standing wave solutions of (1). In , the authors introduce a standing wave ansatz of the following form:(2)ϕt,x=uxeiωt,A0t,x=kx,A1t,x=x2x2hx,A2t,x=-x1x2hx,where ω>0 is a phase frequency and u,k,h are real valued functions on [0,) such that h(0)=0. Inserting (2) into (1), one may check from direct computation that (1) is reduced to the following nonlinear nonlocal elliptic equation:(3)Δu-ωu-xhussu2sdsu-hu2xx2u+λup-2u=0inR2,where hu(s)=0s1/2u2(σ)dσ. See  for its derivation. It is shown in  that (3) is an Euler-Lagrange equation of a C1 functional,(4)Ju12R2u2+ωu2+hu2xx2u2dx-1pR2updxonHr1R2,where Hr1(R2) denotes the set of radially symmetric functions in standard Sobolev space H1(R2). Investigating the structure of J, the authors of  obtain several existence and nonexistence results for (3), depending on the range of p>2 and λ>0. Recently, Pomponio and Ruiz  improve the results in  for the case p(2,4). They find a threshold for the behavior of J, depending on ω>0. They also study (3) on bounded domain in .

In this paper, we are concerned with the existence of infinitely many solutions of (3). It is proved in  that if p>6, J enjoys the symmetric mountain pass geometry and satisfies the (PS) condition so that the well-known symmetric mountain pass lemma (see ) applies to show there exist infinitely many critical points of J. For p(4,6], it turns out that J still enjoys the symmetric mountain pass geometry although checking the (PS) condition is not easy job. One aim of this paper is to show nevertheless J still admits infinitely many critical points for p>4. Moreover, we will replace the power-type nonlinearity λ|u|p-2u of (3) with more general one as follows:(5)Δu-ωu-xhussu2sdsu-hu2xx2u+Vu2u=0.The structure conditions for V are given by the following:

Consider VC1(R+,R) such that V(0)=V(0)=0.

Consider limsupsVs/sp-1<, limsupsVs/sp< for some p>1.

There exists some α>1 such that αV(s)/s2-1/α-V(s)/s3-1/α is monotonically increasing to as s.

Observe that assumptions (V1)–(V3) include the power-type nonlinearity λ/p|u|p, p>4.

Theorem 1.

Assume (V1)(V3). Then, (5) admits infinitely many solutions.

We refer to the work of Cunha et al.  that if we insert a sufficiently small parameter q>0 into (5) as in(6)Δu-ωu-qxhussu2sdsu-qhu2xx2u+Vu2u=0,then much more general assumptions for V, the so-called Berestycki-Lions conditions , are sufficient for guaranteeing the existence and multiplicity of solutions of (6). In our work, we assume further than the Berestycki-Lions conditions but we do not need small parameter q>0. See also  in which Tan and Wan consider asymptotically linear nonlinearities.

To prove Theorem 1, we will apply the method employed in author’s former paper  in which the Schrödinger-Poisson equation, another nonlocal field equation similar to (5), is dealt with. Instead of generating (PS) sequences, we will show the existence and compactness of the so-called approximate solution sequences of J which may be considered as more refined version of (PS) sequences. In Section 2, we give a definition of the approximate solution sequences of J. Some auxiliary lemmas are also prepared in Section 2. In Section 3, we prove the compactness of approximate solution sequences. In Section 4, we construct infinitely many approximate solution sequences whose energy levels go to infinity and complete the proof of Theorem 1.

2. Mathematical Settings and Preliminaries

Let Hr1(R2) be the completion of DuCcR2ux=ux with respect to the norm(7)u=R2u2+ωu2dx1/2.The dual space of Hr1(R2) is denoted by Hr1(R2). Arguing similarly to , it is easy to show (5) is an Euler-Lagrange equation of the C1 functional(8)Ju12R2u2+ωu2+hu2xx2u2-Vu2dxonHr1R2.In this paper, we search for infinitely many critical points of J to prove Theorem 1. To do this, we insert parameter λ into J as follows:(9)Jλu12R2u2+ωu2+hu2xx2u2-λVu2dx.Here λ ranges over [1/2,1]. For a sequence {λj}[1/2,1] which converges to 1 as j, we say {λj,uj} is an approximate solution sequence of J if Jλj(uj)=0 for all j. In the following subsection, we state a variant of the famous Struwe’s monotonicity trick , which plays a crucial role in constructing approximate solution sequences.

2.1. A Variant of Struwe’s Monotonicity Trick

Let X be Banach space. We say a subset AX is symmetric if -vA for every vA. Let A be a compact subset of X and B a closed subset of A. We denote by Γ the set of every continuous odd function γ:AX such that γ(v)=v on B. Let I be a closed interval in R and Jλ one parameter family of even C1 functional on X. We define a minimax level by(10)cλminγΓmaxvAJλγv.The following theorem is a variant of so-called Struwe’s monotonicity trick . A more general version of it is given in . The property (H) below is first proposed by Jeanjean and Toland in .

Theorem 2 (see [<xref ref-type="bibr" rid="B11">11</xref>]).

Suppose that, for all λI,(11)cλ>maxvBJλv.Then, for almost every λI, there exists a norm-bounded (PS) sequence of Jλ at level c(λ), provided the following property (H) for Jλ holds:

For given λ0I, let {λj}I be a sequence strictly increasing to λ0 and {vj} a sequence in X such that(12)-Jλ0vj,Jλjvj,Jλjvj-Jλ0vjλ0-λj

are all uniformly bounded above for j. Then the following holds:

{vj} is norm-bounded in X.

For given ɛ>0, there exists N>0 such that(13)Jλ0vjJλjvj+ɛjN.

2.2. Some Auxiliary Lemmas

Here, we prepare some lemmas which will be necessarily used for proving the main result. Define(14)J^λu12R2hu2xx2u2-λVu2dx.

Lemma 3 (Lemma  3.2 in [<xref ref-type="bibr" rid="B2">3</xref>]).

Let {uj}Hr1(R2) be a sequence weakly converging to some u0 in Hr1(R2) as j. Then, for each φHr1(R2), it holds that J^λ(un)J^λ(u0), J^λ(un)φJ^λ(u0)φ, and J^λ(un)unJ^λ(u0)u0 as j, up to a subsequence.

Lemma 4 (Pohozaev identity).

Let uHr1(R2) be a critical point of Jλ. Then one has(15)R2ωu2dx+2R2hu2xx2u2dx-R2λVu2dx=0.

For each uHr1(R2) and t>0, we define one parameter family of functions utHr1(R2) by(16)utxtαutx.For fixed t>0, we define a map Lt:H1(R2)H1(R2) by Lt(u)=ut. It is easy to see that Lt is a continuous and linear map with the inverse L1/t. Thus Lt is a linear isomorphism.

For each uH1(R2) and λ[1/2,1], let cu,λ:[0,)R be a function defined by(17)cu,λtJλut.

Lemma 5.

For any uHr1(R2) and λ[1/2,1], cu,λ admits a unique critical point t0>0 on (0,); that is, cu,λ(t0)=0, such that cu,λ(t) is increasing on (0,t0), attains its maximum at t0, and is decreasing to - on (t0,).

Proof.

By the change of variable, one can compute (18) c u , λ t = J λ t α u t · = 1 2 t 2 α R 2 u 2 d x + t 2 α - 2 R 2 ω u 2 d x + t 6 α - 4 R 2 h u 2 x x 2 u 2 d x - t - 2 R 2 λ V t 2 α u 2 d x . We differentiate it with respect to t to get (19) c u , λ t = α t 2 α - 1 R 2 u 2 d x + α - 1 t 2 α - 3 R 2 ω u 2 d x + 3 α - 2 t 6 α - 5 R 2 h u 2 x x 2 u 2 d x + t - 3 R 2 λ V t 2 α u 2 d x - t - 2 R 2 λ V t 2 α u 2 α t 2 α - 1 u 2 d x = t 6 α - 5 α t 4 α - 4 R 2 u 2 d x + α - 1 t 4 α - 2 R 2 ω u 2 d x + 3 α - 2 R 2 h u 2 x x 2 u 2 d x A t - R 2 λ α V t α u 2 t α u 2 2 - 1 / α - V t α u 2 t α u 2 3 - 1 / α u 2 3 - 1 / α d x B t . Observe from assumption (V3) that A(t) is strictly monotonically decreasing from infinity to the positive number (3α-2)R2hu2(x)/x2u2dx on (0,) and B(t) is monotonically increasing to infinity on (0,). Therefore there is t0>0 such that A(t)>B(t) on (0,t0), A(t0)=B(t0), and A(t)<B(t) on (0,t0). Also from assumption (V3), we deduce B(t) as t. This proves the proposition.

We define a function T:H1(R2)R by assigning a positive number t satisfying t-αu(t-1·)=1 for any nonzero uHr1(R2). The value T(0) is defined by 0.

Lemma 6.

The function T is well-defined and continuous even map on [0,).

Proof.

To show the well-definedness of T, we have to show that there exists unique t>0 satisfying t-αu(t-1·)=1 for given nonzero uHr1(R2). We note that this is equivalent to prove there is a unique solution t>0 of the equation(20)gtt2α-R2ωu2dxt2-R2u2dx=0.Arguing similarly to the proof of Lemma 5, we are able to see that g(t) is monotonically decreasing on (0,t0) for some t0>0, attains its unique local minimum at t0, and is monotonically increasing to infinity on (t0,). Therefore there is a unique positive zero of g(t) since g(0)=0. Also, the implicit function theorem says that T(u) is continuous on [0,) because g(T(u))0. The evenness of T follows from the fact that each coefficient of (20) is even. This completes the proof.

3. Compactness of Approximate Solution Sequences

In this section, we prove the compactness of an approximate solution sequence {uj} of Jλj when its energy {Jλj(uj)} is bounded above.

Proposition 7.

Let λj[1/2,1] be such that λj1 as j. Let {uj}Hr1(R2) be a sequence of critical points of Jλj; that is, Jλj(uj)=0. Suppose that Jλj(uj)<C for some C>0, independent of j. Then uju0 in H1(R2) for some critical point u0Hr1(R2) of J(=J1) up to a subsequence.

Proof.

We divide the proof into two steps.

Step 1 (boundedness of uj). We first prove that {uj} is bounded in Hr1(R2). Arguing indirectly, suppose that {uj} is unbounded. Let TjT(uj), where the function T is defined in Section 2. Equation (20) says Tj is unbounded. Let vj(x)Tj-αuj(Tj-1x) so that vj=1. Then, up to a subsequence, {vj} converges weakly in Hr1(R2) and strongly in Lq(R2) for all q>2 to some v0Hr1(R2). Since uj is a critical point of Jλj, we see that(21)R2uj2dx+R2ωuj2dx+3R2huj2xx2uj2dx-R2λjVuj2uj2dx=Jλjujuj=0.Combining this with the Pohozaev identity (15), we obtain(22)αR2uj2dx+α-1R2ωuj2dx+3α-2R2huj2xx2uj2dx+R2λjVuj2dx-αR2λjVuj2uj2dx=0.Then, from the change of variable and dividing by Tj6α-4, (22) transforms to(23)αTj4α-4R2vj2dx+α-1Tj4α-2R2ωvj2dx+3α-2R2hvj2xx2vj2dxAj=λjR2αVTjαvj2Tjαvj22-1/α-VTjαvj2Tjαvj23-1/αvj23-1/αdxBj.Since vj=1 and Tj is unbounded, Aj is bounded for j but the structure assumption (V3) implies that Bj tends to infinity as j provided v0 is not identically zero. We claim that v0 is nonzero. Suppose v0 is identically zero. From (19) and (22), we see that(24)cuj,λj1=αR2uj2dx+α-1R2ωuj2dx+3α-2R2huj2xx2uj2dx+R2λjVuj2dx-αR2λjVuj2uj2dx=0.Then, Lemma 5 implies that cuj,λj(1) is the global maximum of cuj,λj(t) on (0,). Thus we see that, for each R>1, (25) C > J λ j u j = c u j , λ j 1 c u j , λ j R T j - 1 = 1 2 R 2 α R 2 v j 2 d x + R 2 α - 2 R 2 ω v j d x + R 6 α - 4 R 2 h v j 2 x x 2 v j 2 d x - R 2 λ j V R α v j R x 2 d x 1 2 R 2 α - 2 v j 2 - 1 2 R 2 λ j V R α v j R x 2 d x = 1 2 R 2 α - 2 + o 1 . The last equality follows from vj=1, the convergence of {vj} to 0 in Lq(R2) for all q>2, and the structure conditions (V1)-(V2). However, taking large R>1, this makes a contradiction and shows v0 is not identically zero. This proves the boundedness of {uj} in Hr1(R2).

Step 2 (compactness of uj). Compactness of {uj} follows from a standard procedure. Since {uj} is bounded, there exists u0Hr1(R2) such that {uj} converges, up to a subsequence, to u0 weakly in Hr1(R2) and strongly in Lq(R2) for all q>2. Then it follows from Lemma 3 that u0 is a critical point of J. Also, it is easy to see from the boundedness of {uj} that Juj0 in Hr1(R2). Recall that(26)Ju=12u2+J^u.Using Lemma 3 once again, one can observe that(27)o1uj=Jujuj=uj2+J^ujuj=uj2+J^u0u0+o1=uj2-u02+o1,which shows uju0 as j. Therefore we have uju0 in Hr1(R2) as j.

4. Construction of Approximate Solution Sequences

In this section, we construct infinitely many approximate solution sequences. Choose an orthonormal basis {ui}i=1 of Hr1(R2). For given kN, let Yk and Zk be linear subspaces of Hr1(R2) spanned by {u1,u2,,uk} and {uk,uk+1,}, respectively. We will show that Jλ enjoys a variant of symmetric mountain pass geometry (see ).

Lemma 8.

There exist a sequence {kn}N such that kn as n and sequences {ρn},{rn}R satisfying ρn>rn>0 for each n and

infJλuuZkn,Tu=rnn for all λ[1/2,1] and all n;

maxJλuuYkn,Tu=ρn0 for all λ[1/2,1] and all n.

Proof.

We first show (i). The structure assumptions (V1)-(V2) imply that(28)Vsω2s+Cspfor someC>0  and alls0.Then,(29)Jλu12R2u2dx+12-λ4R2ωu2dx+12R2hu2xx2u2dx-Cλ2R2u2pdx14R2u2dx+14R2ωu2dx-C2R2u2pdx.We recall the function T and the linear isomorphism Lt in Section 2. Let v(x)T(u)-αu(T(u)-1x) so that v=1. By a change of variable, we get from (29) that, for each uZk,(30)Jλu14Tu2αR2v2dx+14Tu2α-2R2ωv2dx-C2Tu2pα-2R2v2pdx14Tu2α-2-C2Tu2pα-2βk,Tu,where(31)βk,tsupvWk,tR3v2pα-2dx,Wk,t=vHr1R2v=1,vt=tαvt·Zk.We claim that, for each t>0, βk,t0 as k. To see this, suppose that βk,tβt>0 as k. Choose vk,tWk,t satisfying(32)R2vk,tpdx>βk,t2.Since Wk,t is the unit sphere of a linear subspace Lt-1(Zk) of Hr1(R2) with codimension k-1, we deduce {vk,t} converges to 0 weakly in Hr1(R2) and strongly in Lp(R2), up to a subsequence. This however contradicts the fact that(33)limkR2vk,tpdx>βt2>0and the claim is true. We take rn=(8n)1/2α-2. Then, for any uZk satisfying T(u)=rn,(34)Jλu2n-C2rn2pα-2βk,rn.For each n, by taking sufficiently large {kn} satisfying(35)βkn,rn<2nCrn2pα-2,we can see that the proof of (i) is complete.

Next we show (ii). Lemma 5 says that, for each uHr1(R2), Jλ(ut)- as t. Also, we see from Lemma 6 and (20) that the set uYkTu=ρ is closed and bounded in finite dimensional space so it is compact. Combining these two facts with the compactness of [1/2,1], we can deduce easily (ii) holds.

Define(36)BnuYknTuρn,SnuZknTu=rn,with kn,ρn, and rn given in Lemma 8. Let Γn be the set of continuous functions γ:BnHr1(R2) satisfying γ(u)=u on Bn=uBnTu=ρn. By γ(Bn), we denote the set γuHr1R2uBn.

Lemma 9 (intersection property).

For any nN, the intersection γ(Bn)Sn for every γΓn.

Proof.

Choose and fix arbitrary n and γΓn. Define(37)UuintBnTγu<rn,where int(Bn) denotes the interior of Bn in Ykn. Then U is a symmetric open neighborhood of 0 since T is a continuous even map by Lemma 6 and γ is a continuous odd map. Equation (20) in Lemma 6 says Bn is bounded so that U is also bounded. We claim that T(γ(U))=rn. From the continuity of T and γ, it holds that T(γ(U))rn. Suppose that there is some u0U such that T(γ(u0))<rn. Then there is a neighborhood of V of u0 in Ykn such that T(γ(V))<rn. Choose some v0VU. From the definition of U, v0int(Bn). Since v0UBn, we see that v0Bn. Then, from the definition of γ, we have T(γ(v0))=T(v0)=ρn>rn, which is a contradiction. This shows the claim is true.

Now, consider a map Pnγ:UYkn-1, where Pn is the projection map from Hr1(R2)Ykn-1. Then the well-known Borsuk-Ulam theorem applies to see the continuous odd map Pnγ has a vanishing point uU; that is, Pn(γ(u))=0. This means that γ(u)Zkn. Therefore γ(u)γ(Bn)Sn. The proof is complete.

Now, we are ready to prove the existence of infinitely many approximate solution sequences of J. For each λ[1/2,1], we define infinitely many minimax levels as follows:(38)CnλminγΓnmaxuBnJλγu.It follows from Lemmas 8 and 9 that Cn(λ)n for all λ[1/2,1].

Proposition 10.

For every fixed nN, there exists an approximate solution sequence {(λj,uj)} of J such that Jλj(uj)=Cn(λj).

Proof.

We invoke Theorem 2. From Lemmas 8 and 9, it holds that(39)Cnλn>0maxvBnJλv.Let us check Jλ enjoys property (H). Let {λj}[1/2,1] be a sequence strictly increasing to some λ0[1/2,1] and {vj}Hr1(R2) a sequence such that(40)-Jλ0vj,Jλjvj,Jλjvj-Jλ0vjλ0-λj<C.We need to show (i) {vj} is bounded in Hr1(R2) and (ii), for given ɛ>0, there exists N>0 satisfying(41)Jλ0,vjJλj,vj+ɛnN.We first show (i). We see from (40) that(42)12vj2+12R2hvj2xx2vj2dx<2C,-C<12R2Vuj2dx<C,which shows {vj} is bounded in Hr1(R2). Also, for given ɛ>0,(43)Jλ0,vj-Jλj,vj=λj-λ012R2Vuj2dxλ0-λjCɛif j is sufficiently large. This shows (ii). Therefore, there exists a subset M[1/2,1] with full measure in [1/2,1] that, for every λM, there exists a bounded (PS) sequence {uλ,k}k=1 of Jλ at level Cn(λ). Arguing similarly to Step  2 of Proposition 7, we also deduce {uλ,k}k=1 converges, up to a subsequence, to some critical point uλ of Jλ with Jλ(uλ)=Cn(λ). Since M has full measure in [1/2,1], this completes the proof.

Completion of the Proof of Theorem 1. Now we complete the proof of Theorem 1. We first choose and fix arbitrary nN. Let {uj} be an approximate solution sequence of Jλj, given by Proposition 10. Take γΓn satisfying(44)maxuBnJγuCn1+12.It follows from the compactness of Bn that(45)CnλjmaxuBnJλjγumaxuBnJγu+1-λjmaxuBnR212Vγu2dxCn1+12+12=Cn1+1for sufficiently large j. Then Proposition 7 applies to see uj converges, up to a subsequence, to some u0 which is a critical point of J. Recall that Jλj(uj)=Cn(λj)n. By taking a limit j, we deduce J(u0)n. Since n is arbitrary, this shows the existence of infinitely many critical points of J. This completes the proof.

Conflict of Interests

The author declares that there is no conflict of interests regarding the publication of this paper.

Acknowledgments

The author was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (NRF-2014R1A1A2054805) and also was supported by the POSCO TJ Park Science Fellowship.

Jackiw R. Pi S.-Y. Soliton solutions to the gauged nonlinear Schrödinger equation on the plane Physical Review Letters 1990 64 25 2969 2972 10.1103/physrevlett.64.2969 2-s2.0-4243486541 Jackiw R. Pi S.-Y. Classical and quantal nonrelativistic Chern-Simons theory Physical Review D 1990 42 10 3500 3513 10.1103/PhysRevD.42.3500 2-s2.0-25944477679 Byeon J. Huh H. Seok J. Standing waves of nonlinear Schrödinger equations with the gauge field Journal of Functional Analysis 2012 263 6 1575 1608 10.1016/j.jfa.2012.05.024 ZBL1248.35193 2-s2.0-84864029973 Pomponio A. Ruiz D. A variational analysis of a gauged nonlinear Schrödinger equation Journal of the European Mathematical Society 2015 17 6 1463 1486 10.4171/jems/535 Pomponio A. Ruiz D. Boundary concentration of a Gauged nonlinear Schrödinger equation on large balls Calculus of Variations and Partial Differential Equations 2015 53 1-2 289 316 10.1007/s00526-014-0749-2 2-s2.0-84901886028 Huh H. Standing waves of the Schrödinger equation coupled with the Chern-Simons gauge field Journal of Mathematical Physics 2012 53 6 063702 10.1063/1.4726192 2-s2.0-84863524717 Ambrosetti A. Rabinowitz P. H. Dual variational methods in critical point theory and applications Journal of Functional Analysis 1973 14 4 349 381 10.1016/0022-1236(73)90051-7 2-s2.0-34548350707 Cunha P. d'Avenia P. Pomponio A. Siciliano G. A multiplicity result for Chern-Simons-Schrödinger equation with a general nonlinearity Nonlinear Differential Equations and Applications In press Berestycki H. Lions P.-L. Nonlinear scalar field equations. I. Existence of a ground state Archive for Rational Mechanics and Analysis 1983 82 4 313 345 10.1007/bf00250555 2-s2.0-0020591567 Tan J. Wan Y. Standing waves for the Chern-Simons-Schrödinger systems without (AR) condition Journal of Mathematical Analysis and Applications 2014 415 1 422 434 10.1016/j.jmaa.2014.01.084 2-s2.0-84897636919 Seok J. On nonlinear Schrödinger-Poisson equations with general potentials Journal of Mathematical Analysis and Applications 2013 401 2 672 681 10.1016/j.jmaa.2012.12.054 2-s2.0-84873567494 Struwe M. The existence of surfaces of constant mean curvature with free boundaries Acta Mathematica 1988 160 1 19 64 10.1007/bf02392272 2-s2.0-0002725905 Jeanjean L. Toland J. F. Bounded Palais-Smale mountain-pass sequences Comptes Rendus de l'Academie des Sciences—Series I: Mathematics 1998 327 1 23 28 10.1016/s0764-4442(98)80097-9 2-s2.0-0037484917 Willem M. Minimax Theorems 1996 24 Boston, Mass, USA Birkhäuser Progress in Nonlinear Differential Equations and Their Applications